When we deploy redis in kubernetes, each pod corresponds to a redis instance or sentinel instance, using the dns address as the pod's fixed external address, we found:

1. When sentinel is configured with "sentinel announce-ip sentinel-0-svc", sentinel-0-svc is this sentinel`s dns address, sentinel will always report the "+sentinel-address-switch" log when it is started.
2. When slave is configured with "replicaof redis-0-svc", redis-0-svc is the master redis instance`s dns address, sentinel will report the “+slave” log first, followed by the “+fix-slave-config“ log，this will cause the slave to re-initiate replication to the master.
3. When redis slave is configured with "replica-announce-ip redis-0-svc" (redis-0-svc is the dns address of the redis instance) and a failover occurs, sentinel will list the slave instance twice.

So, in these scenarios,

1. When storing redis slave instances in sentinel, ip:port is used as the key uniformly (modify this in line 1198), even if the redis slave uses the dns address to configure replica-announce-ip.
2. It should first convert the dns address (or hostname) to the ip address when search an instance (modify this in line 1397 and line 2372).

This PR is related to #7393